Oblique survival trees based on dipolar splitting criteria

Abstract

Survival analysis, which is the study of time to event distributions, has wide application in epidemiology, engineering and finance, among many others. Semi-parametric and parametric models have been developed to accommodate survival data. More recently, machine learning approaches, such as support vector machines, neural networks and survival forests, have been successfully developed to model survival data. Ensemble methods such as survival forests depend on splitting data at nodes in underlying decision trees. Various splitting criteria have been proposed and implemented using within or between-node homogeneity. In this work, we show improvement and clarification of existing algorithms which rely on non-parametric, dipolar splits by hyperplanes for maximizing between-node homogeneity. We will demonstrate the predictive power of these decision trees on real and simulated data sets. These models can be used in ensemble methods to reduce variability and improve predictive power on test sets.